materials

Article
Discrete Element Method Modeling of the
Rheological Properties of Coke/Pitch Mixtures
Behzad Majidi 1,2 , Seyed Mohammad Taghavi 2 , Mario Fafard 2 , Donald P. Ziegler 3 and
Houshang Alamdari 1,2, *
1 Department of Mining, Metallurgical and Materials Engineering, Laval University,
Quebec, QC G1V0A6, Canada; behzad.majidi.1@ulaval.ca
2 NSERC/Alcoa Industrial Research Chair MACE and Aluminum Research Center, Laval University,
Quebec, QC G1V0A6, Canada; Seyed-Mohammad.Taghavi@gch.ulaval.ca (S.M.T.);
Mario.Fafard@gci.ulaval.ca (M.F.)
3 Alcoa Primary Metals, Alcoa Technical Center, 100 Technical Drive, Alcoa Center,
New Kensington, PA 15069-0001, USA; donald.ziegler@alcoa.com
* Correspondence: houshang.alamdari@gmn.ulaval.ca; Tel.: +1-418-656-7666

Academic Editor: Martin O. Steinhauser

Received: 2 February 2016; Accepted: 28 April 2016; Published: 4 May 2016

Abstract: Rheological properties of pitch and pitch/coke mixtures at temperatures around 150 C are
of great interest for the carbon anode manufacturing process in the aluminum industry. In the present
work, a cohesive viscoelastic contact model based on Burgers model is developed using the discrete
element method (DEM) on the YADE, the open-source DEM software. A dynamic shear rheometer
(DSR) is used to measure the viscoelastic properties of pitch at 150 C. The experimental data obtained
is then used to estimate the Burgers model parameters and calibrate the DEM model. The DSR tests
were then simulated by a three-dimensional model. Very good agreement was observed between the
experimental data and simulation results. Coke aggregates were modeled by overlapping spheres in
the DEM model. Coke/pitch mixtures were numerically created by adding 5, 10, 20, and 30 percent
of coke aggregates of the size range of 0.2970.595 mm (30 + 50 mesh) to pitch. Adding up to 30%
of coke aggregates to pitch can increase its complex shear modulus at 60 Hz from 273 Pa to 1557 Pa.
Results also showed that adding coke particles increases both storage and loss moduli, while it does
not have a meaningful effect on the phase angle of pitch.

Keywords: discrete element method; viscoelastic; particles; dynamic shear rheometer;

simulation; pitch

1. Introduction
Carbon anodes for the aluminum smelting process are made using a well-designed recipe of
pitch binder and calcined coke particles. Anodes are consumed during the electrolysis process in
Hall-Hroult cells, and it is estimated that around 17% of the cost of the aluminum smelting process
comes from carbon anodes [1]. Thus, the properties and quality of the anodes have a direct effect on
the performance and economy of the Hall-Hroult process. Anodes are made by mixing granulated
coke aggregates with coal tar or petroleum pitch at 150 C and the mixture, known as anode paste,
is then compacted and baked. It is essential to have a good understanding of the viscosity and
rheological properties of pitch and pitch/coke mixtures to have better control of final compacted
anode properties. Micromechanical models can provide considerable information about the flow and
compaction properties of anode paste.
Discrete element method (DEM) is used to simulate the behavior of granular materials in
industrial applications especially where the dynamics and flow of a particulate material are of interest.

Materials 2016, 9, 334; doi:10.3390/ma9050334 www.mdpi.com/journal/materials

Materials 2016, 9, 334 2 of 13

This method was introduced for the first time by Cundall and Strack [2] in 1979. In DEM simulations,
rigid discrete elements, which are spheres in 3D and discs in 2D models, are used to model the granular
material. The contact law between the elements defines the mechanical behavior of the bulk material.
The discrete element approach has attracted the interest of researchers in mining, civil engineering,
pharmaceutical industries, and materials engineering to simulate the flow [3,4], compaction [5,6],
and mechanical properties [7,8] of single- and multi-phase materials. Recent advancements in the
performance and power of computers have provided new insights to simulate the mechanical and
rheological properties of asphalt concretes and mastics using two- and three-dimensional DEM models.
For instance, Dondi et al. [9] studied the effects of aggregate size and shape on the performance
of asphalt mixes. Khattak et al. [10] used an imaging technique coupled with DEM to create
a micro-mechanical model of hot mix asphalts (HMA). They reported that the DEM model provided
a good prediction of the dynamic modulus and strength of HMAs. Since materials properties in DEM
are modeled by assigning appropriate contact models to the elements, different elastic, elasto-plastic,
and viscoelastic models are developed. For asphalt mixes, usually Burgers viscoelastic model is used
for the mastic, and a simple elastic model is used for the aggregates [8,11]. Ma et al. [11] used Burgers
model in a three-dimensional DEM model to investigate the effects of air voids on creep behavior of
asphalt mixtures.
Burgers model can be also embedded in traditional linear elastic contact models in which normal
and shear stiffness of the contact changes by time to include viscous deformations. Vignali et al. [12]
and Dondi et al. [13] adapted this approach of viscoelastic modeling to predict the rheological behavior
of bituminous binders.
In the present work, three-dimensional imaging is used to capture the irregular shapes of coke
aggregates. Then, a 3D DEM model of pitch is developed to predict the rheological properties of pitch
and coke/pitch mixtures at 150 C.

2. Theory
A three-dimensional DEM model is composed of a combination of discrete spheres and walls.
At the beginning, the position of all elements and walls are known so that the active contacts are easily
determined. Then, according to the mechanical behavior of the material, an appropriate force-displacement
law is applied to each contact, and the contact forces are calculated. Newtons second law of motion is
then used to update the position and velocity of each element.
One common contact model widely used in DEM simulations is the linear contact model.
This model is simply defined by assigning normal and shear stiffness values to the contacting elements.
Contact forces can be calculated according to the extent of the overlap (for normal contact force) and
tangent movement (for shear contact force):

Fn Kn U n (1)

F s K s U s (2)

Irregularly shaped particles can be generated as a clump composed of several touching or

overlapping balls. Contact force calculations for balls within a clump is skipped during calculation
cycle and only the contacts of a clump with neighboring clump/balls or walls are considered.
Burgers four-element model is the most common model used to simulate the viscoelastic
properties of mastics and binders [14]. Burgers model, as shown in Figure 1, is composed of the
Maxwell model in series with the Kelvin model; thus, it is capable of representing the material behavior
under both creep and relaxation [14].
 Figure 1. Burgers four-element model composed of the Maxwell (m) model in series with the Kelvin
Materials 2016, 9, 334 3 of 13
(k) model.
Materials 2016, 9, 334 3 of 12

Deformation of Burgers model is the sum of the deformations of Kelvins elements and
Maxwells elements. This means that the total deformation can be written as:
= + (3)
Deformation of the Maxwell element, in turn, comes from the dashpot and the spring:
Figure 1. Burgers four-element model composed of the Maxwell (m) model in series with the Kelvin
Figure 1. Burgers(k)four-element model composed of the Maxwell (m) model in series with the Kelvin
= + +
model.
(4)
(k) model.
Deformation of Burgers model is the sum of the deformations of Kelvins elements and
The first andMaxwells
secondelements.
derivatives ofthat
This means Equation (4) cancan
the total deformation bebewritten
written as: as:

Deformation of Burgers model is the sum =the+deformations of Kelvins elements (3)

= of + + and Maxwells
(5)
Deformation of the Maxwell element, in turn, comes
elements. This means that the total deformation can be written as: from the dashpot and the spring:

= = ++ + + (4) (6)
The first and second derivatives of Equation
u u(4) `canube written as: (3)
The force-displacement equation of Burgers kmodel m can be written as:
= + + (5)

Deformation of the Maxwell 1 in1 =turn,

+ comes
+
+ element, from
= the dashpot and the spring:
(6)
+ + + (7)
The force-displacement equation of Burgers model can be written as:
1 u1 ` u
Although Burgers model is widely + + uused + by k some + mk ` u=mc
researchers, mostly to simulate (7) the mechanical (4)
behavior of asphalt mastic, it hasmodel
Although Burgers notisbeen publicly
widely used by some implemented
researchers, mostly toin YADE
simulate [15]. Therefore, authors
the mechanical
The first and behavior
secondof derivatives
asphalt mastic, it ofhas Equation
not been publicly (4)implemented
can be written in YADE as: [15]. Therefore, authors
use an in-house implementation of this model in YADE, creating a
use an in-house implementation of this model in YADE, creating a cohesive viscoelastic model to
cohesive viscoelastic model to
simulate pitch. A new type of material, . .
theu so-called . .
simulate pitch. A new type of material,
u
material takes four of Burger's parameters in
the
k`
so-called
` uCohBurgersMat
CohBurgersMat
umkdirection
normal
is defined inis defined
YADE.
mcand four of Burgers parameters in
This in YADE. This (5)
material takes four of Burger's parameters in normal direction and
shear direction. Interactions of elements having CohBurgersMat are governed by Equation (7).
four of Burgers parameters in
.. .. .. ..
shear direction. Interactions
Viscoelastic of elements
properties
uhaving
of a material
are CohBurgersMat
uk normally
` umk ` presented by two
umc are governed
parameters by Equation (7). (6)
of a complex
shear modulus, G*, and phase angle, [13]. These parameters can be measured using a parallel-plate
Viscoelastic properties
measuring system ofasaschematically
materialshown are innormally
Figure 2. Thepresented by two(DSR)
dynamic shear rheometer parameters
shown of a complex
shearThe force-displacement
modulus, G*, and 3a
in Figure phase equation
is the of
angle, equipment
laboratory Burgers
[13]. widelymodel
These parameters can becan
used to characterize written
thebe as: properties
measured
rheological using
of a parallel-plate
different types of binders and mastics [16]. As shown in Figures 2 and 3b, a disc of the tested material
measuring systemis sandwiched
as schematically
between twoplates
Ck shown
1 at the1in
Figure
desired 2. CTheIn this
.temperature.
C dynamic shear
test, a sinusoidal
. Crheometer
Ckforce (stress) (DSR) shown
k m m ..
f
in Figure 3a is the laboratory` ` C ` f ` f C u u
theK rheological properties(7)
Kk lagequipment Kkfrom widely used to characterize of
is applied to the sample, mand the deformation (strain) is recorded. The induced m strain has a sinusoidal
form with a time that comes K Kk K
themmaterials viscose m
deformation. This time lag isk called the
different types of phase
binders and mastics [16]. As shown in Figures 2 and 3b,
angle (). The frequency sweep configuration is adopted to obtain the response of fluid-like a disc of the tested material
is sandwiched materials model
between
Although Burgers to
twodifferentisloading
plates atfrequencies
widely theused by
desired
by measuring the
temperature.
some complex shear
researchers, modulus
In this
mostly and
test, toaphase angle
sinusoidal
simulate theforce (stress)
mechanical
at different frequencies.
is appliedoftoasphalt
behavior the sample,
mastic, and the deformation
it has not been publicly (strain) is recorded.inThe
implemented YADE induced strain has aauthors
[15]. Therefore, sinusoidal
use
form
an with aimplementation
in-house time lag that comes of this frommodel thein materials
YADE, creating viscoseadeformation.
cohesive viscoelastic This time lag istocalled
model the
simulate
phase A
pitch. angle
new().
typeThe frequencythe
of material, sweep configuration
so-called CohBurgersMat is adopted to obtain
is defined the response
in YADE. of fluid-like
This material takes
materials
four to different
of Burger's loadinginfrequencies
parameters normal direction by measuring and four theofcomplex
Burgersshear modulus
parameters inand
shearphase angle
direction.
at different frequencies.
Interactions of elements having CohBurgersMat are governed by Equation (7).

Figure 2. Schematic illustration of a dynamic shear rheometer.

Figure 2.
Figure 2. Schematic
Schematic illustration
illustration of
of aa dynamic
dynamic shear
shear rheometer.
rheometer.

Viscoelastic properties of a material are normally presented by two parameters of a complex

shear modulus, G*, and phase angle, [13]. These parameters can be measured using a parallel-plate
measuring system as schematically shown in Figure 2. The dynamic shear rheometer (DSR) shown
in Figure 3a is the laboratory equipment widely used to characterize the rheological properties of
different types of binders and mastics [16]. As shown in Figures 2 and 3b, a disc of the tested material
is sandwiched between two plates at the desired temperature. In this test, a sinusoidal force (stress) is
Materials 2016, 9, 334 4 of 13

applied to the sample, and the deformation (strain) is recorded. The induced strain has a sinusoidal
form with a time lag that comes from the materials viscose deformation. This time lag is called the
phase angle (). The frequency sweep configuration is adopted to obtain the response of fluid-like
materials to different loading frequencies by measuring the complex shear modulus and phase angle
at different frequencies.
Materials 2016, 9, 334 4 of 12

(a) (b)
Figure 3.
Figure 3. Dynamic
Dynamic shear
shear rheometer
rheometer (DSR)
(DSR) (a);
(a); and
and aa pitch
pitch sample
sample being
being tested
tested (b).
(b). The
The black
black part
part in
in
the DSR instrument is the sample bath which keeps the temperature constant during the test.
the DSR instrument is the sample bath which keeps the temperature constant during the test.

The following equations are used to calculate G*:

The following equations are used to calculate G*:
2
= (8)

2T
max 3 (8)
r
= (9)

r
(9)
in which is the maximum shear stress, maxis theh maximum applied torque, is the radius of
thewhich
in max is the
specimen, is the rotationshear
maximum angle,stress,
and Tisis the
the thickness
maximumofapplied
the specimen.
torque, Finally, the complex
r is the radius of the
shear modulus
specimen, is obtained
is the rotation from:
angle, and h is the thickness of the specimen. Finally, the complex shear
modulus is obtained from:

= max (10)
G (10)
max
The phase
The phase angle
angle (),
(), in
in turn,
turn, was
was determined
determined at at different
different frequencies
frequencies by
by measuring
measuring the the delay
delay in
in
seconds between the peaks in the stress and strain functions.
seconds between the peaks in the stress and strain functions.
Having obtained G and
Having obtained and ,, the
the storage
storage and
and loss
loss moduli
moduli atat 150
150 C
C for
for each frequency are
each frequency are
calculated by:
calculated by:
G1
= G
cos
cos (11)
(11)
= G

G2" sin
sin (12)
(12)

3. Experimental
3. Experimental
Dynamic shear tests were conducted in strain-controlled mode using an ARES rheometer (see
FigureDynamic shear sweep
3). Frequency tests were conducted
configuration in adopted
was strain-controlled
at 150 C mode using
from 0.06 Hz an ARES
to 60 rheometershear
Hz. Complex (see
Figure 3). Frequency sweep configuration was adopted at 150 C from 0.06 Hz
modulus and phase angle were measured at each frequency. Then, the obtained data were used to to 60 Hz. Complex
shear modulus
estimate the fourand phase angle
parameters were measured
of Burgers model atof each
pitchfrequency.
at 150 C. Then, the obtained data were used
to estimate the four parameters of Burgers model of pitch at 150 C.
Numerical Method
Numerical Method
The open-source discrete element code, YADE [15], was used in this work to simulate the DSR test.
The open-source
The geometry discrete model
of the numerical elementofcode, YADE
the DSR test [15], wasisused
of pitch shown in is
this work4.toPitch
Figure simulate the DSR
is modeled by
test. The geometry of the numerical model of the DSR test of pitch is shown
an assembly of spheres with the radius of 0.08 mm. Spheres are generated in hexagonal closed is Figure 4. Pitch
packis
modeled by antoassembly
configuration make upofa spheres with
disc of 12 mmthein radius
diameterof 0.08
andmm.
2 mm Spheres are generated
in thickness. in hexagonal
Size of the elements
closed pack configuration to make up a disc of 12 mm in diameter and 2 mm in thickness. Size of the
elements has been chosen considering the balance between the resolution of the model (size of the
elements compared to the size of the sample) and the computation time.
Materials 2016, 9, 334 5 of 13

has been chosen considering the balance between the resolution of the model (size of the elements
compared to9,the
Materials 2016, 334 size of the sample) and the computation time. 5 of 12

Figure 4.
Figure 4. Discrete element method
Discrete element method (DEM)
(DEM) simulation
simulation of
of the
the DSR
DSR test
test of
of pitch.
pitch.

AA small
small compressive
compressive load load waswas initially
initially applied
applied to to ensure
ensure thethe contacts
contacts between
between the
the plates
plates and
and the
the
testing material.
testing material. Frequency
Frequency sweep sweep dynamic
dynamic shear shear tests
tests were
were run
run atat frequencies
frequencies ranging
ranging from
from 0.06
0.06 Hz
Hz
to 60 Hz, according to the DSR
to 60 Hz, according to the DSR experiments. experiments.
Equations (8)(12)
Equations (8)(12) were
were then
then used
used to to calculate
calculate thethe rheological
rheological parameters.
parameters. A A cohesive
cohesive contact
contact
model based on Burgers model (Equation (7)) is used to simulate the
model based on Burgers model (Equation (7)) is used to simulate the viscoelastic behavior of pitchviscoelastic behavior of pitch
at 150 C.
The spherical elements representing pitch have Burgers model
at 150 C. The spherical elements representing pitch have Burgers model in both shear and normal in both shear and normal
direction in
direction in all
all contacts.
contacts. In In this
this work,
work, the the same
same values
values were
were used
used for
for Burgers
Burgers model
model ofof pitch
pitch in
in both
both
shear and normal directions.
shear and normal directions.
In 2011,
In 2011, Liu
Liu et
et al.
al. [17]
[17] proposed
proposed aa new new method
method in in using
using the
the frequency-temperature
frequency-temperature superposition
superposition
(FTS). Traditionally,
(FTS). Traditionally,after afterconducting
conductingdynamic dynamicmodulus
modulusteststestsatat different
different temperatures,
temperatures, FTSFTS is used
is used in
in calculating shift factors to build up the master curve at the reference temperature.
calculating shift factors to build up the master curve at the reference temperature. However, in the However, in the

approach proposed
approach proposedbyby LiuLiu
et al.,etGal.,and and are measured
are measured at different atfrequencies
different frequencies at desired
at desired temperatures
temperatures in the laboratory and the Burgers model parameters are
in the laboratory and the Burgers model parameters are obtained. Then, Burgers dashpot viscosities obtained. Then, Burgers

dashpot viscosities are modified
to predict
are modified to predict G and at amplified frequencies. This method is of considerable interestisfor
and at amplified frequencies. This method of
considerable
discrete element interest for discrete
simulations element by
as for example simulations as for example
using the amplifying by using
coefficient of the amplifying
1000, dynamic
coefficient of = 1000, dynamic modulus test at the frequency of 0.06
modulus test at the frequency of 0.06 Hz can be simulated at f = 60 Hz, resulting in significantly Hz can be simulated at f = 60
Hz, resulting
reduced in significantly
computation time. Inreduced
the present computation
work, time.1000In theused
was present DSR tests
work,
in the = 1000 was used in
simulations.
the DSR tests simulations.
4. Results and Discussion
4. Results and Discussion
4.1. DSR of Pitch and Model Verification
4.1. DSR of Pitch anddata
Experimental Model Verification
of the DSR tests of pitch at 150 C are given in Figure 5. The obtained data
is then used to fit the
Experimental Burgers
data model
of the DSR parameters.
tests TheC
of pitch at 150 fitting procedure
are given is performed
in Figure by the data
5. The obtained solver
is
option in Microsoft Excel by minimizing the following function:
then used to fit the Burgers model parameters. The fitting procedure is performed by the solver
option in Microsoft Excel by minimizing the following
ff2 function: ff2
m
G1 j G2 j
f p
10
1 ` " 1 q (13)
=
j1 (G j 1 +
G2 0j 1 ) (13)
"
1 0 20
where G j and
where and G "j are are respectively
respectively the
the storage
storage and
and loss
loss moduli
moduli experimentally
experimentally measured
measured atat jth
jth
frequency; G 1 and G 2 are respectively the predicted storage and loss moduli at jth frequency; and m
frequency; j and " are respectively the predicted storage and loss moduli at jth frequency; and
j
is the number of data points, which is
m is the number of data points, which is 16.16.

Figure 5. Dynamic shear tests results of pitch at 150 C.

 (13)
"

where and " are respectively the storage and loss moduli experimentally measured at jth
frequency; and " are respectively the predicted storage and loss moduli at jth frequency; and
Materials 2016, 9, 334 6 of 13
m is the number of data points, which is 16.

Figure 5. Dynamic
Figure 5. Dynamicshear
shear tests results
tests results of of pitch
pitch at 150
at 150 C. C.

Calculated values for Burgers model of pitch have been presented in Table 1. These parameters
are called macroscale parameters, and they determine the global material behavior. The Burgers
model parameter values for each contact in the DEM model of pitch, however, depend on the size
of two contacting spheres. In our implementation of Burgers model on YADE, these parameters are
obtained by the following equation for each contact:

Pm L PM (14)

in which PM is the macroscale parameter of Burgers model as given in Table 1, L is the diameter of
the element, and Pm is the microscale parameters of Burgers model of pitch. For a contact formed by
two overlapping elements, however, the contact model parameters are obtained as:

P1 P2
Pcontact 2 (15)
P1 ` P2

Table 1. Calculated Burgers model parameters of pitch at 150 C.

Km (Pa) Cm (Pa.s) Kk (Pa) Ck (Pa.s)

3867.136 37.111 10.047 7.375

In the DEM code here, as a contact is formed, its rheological parameters are obtained from
Equation (15) and then the force-displacement law, Equation (7), is solved at each cycle as long as the
contact is active.
Using the values obtained for pitch and the configuration previously explained in Section 3, DSR
tests were simulated in YADE. The DEM model of the DSR test was calibrated by changing the stiffness
of the loading plate. Stiffness of the plate was modified to obtain the complex shear modulus of pitch at
f = 60 Hz. Then, this value was kept constant in DSR tests at all other frequencies. Using this method,
stiffness of the loading plate was determined as 40,000 N/m.
Figure 6 presents the simulation results of a DSR test of pitch in terms of stress and strain curves
by which, using previously mentioned equations, G and are calculated. Experimentally measured
rheological data (G*, G', and G") of pitch at 150 C in Figures 79 are compared with those obtained by
the DEM simulation with the viscoelastic parameters given in Table 1.
of pitch at f = 60 Hz. Then, this value was kept constant in DSR tests at all other frequencies. Using
of pitch at f = 60 Hz. Then, this value was kept constant in DSR tests at all other frequencies. Using
this method, stiffness of the loading plate was determined as 40,000 N/m.
this method, stiffness of the loading plate was determined as 40,000 N/m.
Figure 6 presents the simulation results of a DSR test of pitch in terms of stress and strain curves
Figure 6 presents the simulation results of a DSR test of pitch in terms of stress and strain curves
by which, using previously mentioned equations, and are calculated. Experimentally measured
by which, using previously mentioned equations, and are calculated. Experimentally measured
rheological
Materials 2016, data
9, 334(G
*, G', and G") of pitch at 150 C in Figures 79 are compared with those obtained
7 of 13
rheological data (G*, G', and G") of pitch at 150 C in Figures 79 are compared with those obtained
by the DEM simulation with the viscoelastic parameters given in Table 1.
by the DEM simulation with the viscoelastic parameters given in Table 1.

Figure 6.
6. DEM results
results as strain
strain and stress
stress curves of
of pitch at
at f = 60 Hz.
Figure 6. DEM
Figure DEM results as
as strain and
and stresscurves pitch atff == 60
curves ofpitch 60 Hz.
Hz.

Figure 7. Experimental and DEM simulation results for complex modulus of pitch at 150 C.
Figure 7. 7. Experimental and DEM simulation results for complex modulus of pitch at 150 C.
Materials 2016, 9, 334 Experimental and DEM simulation results for complex modulus of pitch at 150 C.
Figure 7 of 12
Materials 2016, 9, 334 7 of 12

Figure 8.
8. Experimental and
and DEM
DEM simulation
simulation results
results for
for storage
storage modulus
modulus ofpitch
pitch at150
150C.
Figure Experimental modulus of
of pitch at C.
at 150 C.

Figure 9. Experimental and DEM simulation results for loss modulus of pitch at 150C.
Figure 9.
Figure 9. Experimental and DEM
Experimental and DEM simulation
simulation results
results for
for loss
loss modulus
modulus of
of pitch
pitch at
at 150
150 C.
C.

As can
As can bebe seen
seen from
from Figures
Figures 79,
79, the
the DEM
DEM model
model ofof pitch
pitch is
is capable
capable of
of predicting
predicting the
the complex
complex
modulus at
modulus at aa wide
wide range
range of
of frequencies
frequencies with
with aa very
very good
good precision.
precision. However,
However, storage
storage modulus
modulus at at
high frequencies is over-estimated. This comes from the under-estimation of phase angle
high frequencies is over-estimated. This comes from the under-estimation of phase angle at higher at higher
frequencies. Similar
frequencies. Similar observation
observation has
has been
been reported
reported byby Liu
Liu and
and You
You [17].
[17]. They
They reported
reported that
that the
the errors
errors
in predictions
in predictions ofof the
the complex
complex modulus
modulus andand phase
phase angle
angle for
for the
the amplified
amplified frequencies
frequencies of
of less
less than
than
Materials 2016, 9, 334 8 of 13

As can be seen from Figures 79 the DEM model of pitch is capable of predicting the complex
modulus at a wide range of frequencies with a very good precision. However, storage modulus at
high frequencies is over-estimated. This comes from the under-estimation of phase angle at higher
frequencies. Similar observation has been reported by Liu and You [17]. They reported that the errors
in predictions of the complex modulus and phase angle for the amplified frequencies of less than
25 kHz is around 4%. However, it increases up to 15% for higher frequencies. Using a smaller time-step
and not applying the FTS to avoid its probable inertial effects can improve the predictions of the
phase angle.

4.2. DEM Simulation of Coke/Pitch Pastes

The verified discrete element model of pitch is then used to predict the viscoelastic properties of
pitch/coke mixtures at 150 C. It should be noted that DSR test measurements on coke/pitch mixtures
are not always possible or precise due to the small thickness of the sample disc in the DSR tests and
limitations of the setup. Thus, a well-designed DEM model is of considerable practical interest.
Calcined coke has particles with irregular shapes and, as used in anodes, a wide size distribution.
For discrete element models in the present work, the method previously developed by the authors [18]
was applied to model the coke aggregates by overlapping spheres. Coke particles of the size range of
30 + 50 mesh (0.2970.595 mm) were modeled as clumps in YADE. An example of a modeled coke
particle is shown
Materials in Figure 10.
2016, 9, 334 8 of 12

Figure 10. 3D image of a coke particle and its twin DEM model.
Figure 10. 3D image of a coke particle and its twin DEM model.

Four numerical mixes of coke and pitch with 5, 10, 20, and 30 wt % of coke aggregates were
Four numerical
generated. Then, themixes
DSR testof was
cokesimulated
and pitch with
on the 5, 10,with
samples 20,theand
same30configuration
wt % of coke aggregates
of the only-
werepitch
generated. Then, the DSR test was simulated on the samples with
sample. The DEM model of coke/pitch mixtures is composed of two phases: spheres in the same configuration
of the only-pitch
hexagonal sample.
closed packedThe (hcp)DEM model ofrepresenting
configuration coke/pitch pitch
mixtures
and is composed
clumps of twocoke
representing phases:
aggregates.
spheres in hexagonal closed packed (hcp) configuration representing pitch and clumps representing
To create the coke/pitch sample models, first, clumps of coke aggregates with the intended size
coke aggregates.
distribution andcoke/pitch
To create the numbers are randomly
sample placed
models, in aclumps
first, disc withofa coke
radius of 6 mm and
aggregates a height
with of 2 mm. size
the intended
The model is cycled to remove the possible overlaps between aggregates. At this step, zero gravity is
distribution and numbers are randomly placed in a disc with a radius of 6 mm and a height of 2 mm.
used in the model to prevent the particles from sinking to the bottom of the cylindrical container.
The model is cycled to remove the possible overlaps between aggregates. At this step, zero gravity
Then, using spheres with radii of 0.08 mm, the sample volume is tessellated in an hcp configuration.
is used in the model to prevent the particles from sinking to the bottom of the cylindrical container.
Since there are already clumps in the space, there will obviously be some overlaps between the
Then, using
clumps andspheres with
spheres, radiimust
which of 0.08 mm, the to
be managed sample
have avolume is tessellated in an hcp configuration.
stable sample.
Since there
Using are already
a Python script,clumps in the space,
big clump-sphere thereare
overlaps will obviously
removed. Thebe some
code overlaps
detects between the
the interactions
clumps and spheres, which must be managed to have a stable sample.
in which the penetration depth is larger than 0.02 mm and deletes the standalone sphere (pitch
Using a Python
element). However,script,
if thebig clump-sphere
overlap overlaps
value is less are removed.
than 0.02 mm, the size Theofcode detects thesphere
the standalone interactions
is
in which theto
reduced penetration depth is
0.04 mm without larger it.
deleting than
An0.02 mm and
example of a deletes
result ofthe
thisstandalone spherein
method is shown (pitch element).
Figure 11.
However, if the overlap value is less than 0.02 mm, the size of the standalone sphere is reduced to
0.04 mm without deleting it. An example of a result of this method is shown in Figure 11.
 Since there are already clumps in the space, there will obviously be some overlaps between the
clumps and spheres, which must be managed to have a stable sample.
Using a Python script, big clump-sphere overlaps are removed. The code detects the interactions
in which the penetration depth is larger than 0.02 mm and deletes the standalone sphere (pitch
element).
Materials However,
2016, 9, 334 if the overlap value is less than 0.02 mm, the size of the standalone sphere9 of
is 13
reduced to 0.04 mm without deleting it. An example of a result of this method is shown in Figure 11.

Figure 11. Coke/pitch mixture generation method. (a) Creating coke aggregates; (b) adding pitch
Figure 11. Coke/pitch mixture generation method. (a) Creating coke aggregates; (b) adding pitch
spheres and handling the overlaps.
spheres
Materials and
2016, 9, 334handling the overlaps. 9 of 12

Figure 11a shows the first step in which only clumps are generated. In Figure 11b, pitch spheres
have been added, and the overlapping spheres have been deleted or reduced in diameter to avoid
huge overlaps. The
The resulting
resulting paste
paste then
then undergoes
undergoes aa triaxial
triaxial compression
compression to have an integrated
compacted sample.
In Figure 12, the numerical sample
sample of
of pitch
pitch and
and coke
coke mixture
mixture ready
ready for
for the
the DSR
DSR tests
tests is
is shown.
shown.
The dimensions and test procedure were the same as those for pitch. Four pastes of coke and pitch
mixture were numerically prepared with 5, 10, 20, and 30 wt % of coke particles. Size distribution of
coke aggregates was the same for all samples in the
the range
range of
of 30
30 ++50
50mesh.
mesh.

Figure 12.
Figure 12. Dynamic shear
shear test
test of
of coke/pitch
coke/pitch mixture.
mixture.

The effect of the content of coke particles on the complex shear modulus of coke/pitch pastes has
The effect of the content of coke particles on the complex shear modulus of coke/pitch pastes has
been shown in Figure 13. As the content of coke in the mixture increases, the complex shear modulus
been shown in Figure 13. As the content of coke in the mixture increases, the complex shear modulus
increases. This raise in is more pronounced for higher frequencies. However, the addition of coke
increases. This raise in G is more pronounced for higher frequencies. However, the addition of
particles does not have a meaningful effect on the value phase angle and, as the content of the coke
coke particles does not have a meaningful effect on the value phase angle and, as the content of the
particles increases, both storage and loss moduli increase. Similar results have been reported by
coke particles increases, both storage and loss moduli increase. Similar results have been reported by
Pasquino [19] for glass bead suspensions in viscoelastic polymers.
Pasquino [19] for glass bead suspensions in viscoelastic polymers.
 The effect of the content of coke particles on the complex shear modulus of coke/pitch pastes has
been shown in Figure 13. As the content of coke in the mixture increases, the complex shear modulus
increases. This raise in is more pronounced for higher frequencies. However, the addition of coke
particles does not have a meaningful effect on the value phase angle and, as the content of the coke
particles
Materials 2016,increases,
9, 334 both storage and loss moduli increase. Similar results have been reported10by of 13
Pasquino [19] for glass bead suspensions in viscoelastic polymers.

Figure 13. Effect of coke aggregate content on complex shear modulus of coke/pitch mixtures.
Figure 13. Effect of coke aggregate content on complex shear modulus of coke/pitch mixtures.

Figures 1315 show that there is a considerable rise in the dynamic moduli of pitch by adding coke
particles. In1315
Figures Figure show that there
14, results have is also
a considerable
been compared rise in thetwo
with dynamic moduli
equations of pitchfor
developed by the
adding
case coke
of
particles.
spherical rigid particles in Newtonian fluids, the Hashin-Shtrikman [20] and Krieger-Doughertycase
In Figure 14, results have also been compared with two equations developed for the [21] of
spherical rigid
equations, particles
which in Newtonian
are respectively fluids,
written as: the Hashin-Shtrikman [20] and Krieger-Dougherty [21]
equations, which are respectively written as:
() 2 + 3
= (16)

G1 pq 22
`23
(16)
()G1 0 2 2
= (1 1.5625) . (17)

G1 pq
where is the volume content of particles. 1 p1 1.5625q1.6 (17)
( )
G 0
In Figure 14, the ratio for pitch mixed coke particles are larger than those predicted by the
where is the volume content of particles.
equations. It shouldGbe 1pnoted
q that these equations were proposed for Newtonian fluids including
In Figure 14, the G10 ratio for pitch mixed coke particles are larger than those predicted by the
Materials 2016, 9, 334 10 of 12
equations. It should be noted that these equations were proposed for Newtonian fluids including rigid
spherical particles.
rigid spherical However,
particles. pitch ispitch
However, a non-Newtonian
is a non-Newtonian viscoelastic fluid fluid
viscoelastic above its softening
above point.
its softening
Added particles here are not spherical either, and they are relatively coarse. Therefore,
point. Added particles here are not spherical either, and they are relatively coarse. Therefore, the case the case seems
complex, and we have
seems complex, and we foundhavenofound
analytical or empirical
no analytical equation
or empirical to describe
equation the effects
to describe of adding
the effects of
rigid particles
adding rigid to a viscoelastic
particles fluid. Performing
to a viscoelastic fluid. Performing experimental
experimentalmeasurement
measurement waswas
notnotpossible
possibledue
to the
due size
to the ofsize
theofparticles (in the
the particles (in range
the range of 0.2970.595
of 0.2970.595 mm)
mm)and andthethe stiffness ofthe
stiffness of thesample,
sample, which
which
exceeded
exceeded thethe
limits
limitsofof
our
oursetup.
setup.Therefore,
Therefore,aa mix mix of pitch with withfine
fineparticles
particlessmaller
smallerthan
than 150
150 micron
micron
waswas
mademade forfor experimental
experimental measurements.
measurements. TheThe results
results ofof
DSRDSR tests
tests onon these
these samplescompared
samples comparedtotothe
the above-mentioned
above-mentioned equations
equations are in
are given given in Figure
Figure 16. Again,16. Again,
it can beit can
seenbethat
seen that
the the experimental
experimental point of
G1 of the
point ratio for pitch/coke mixtures are larger than the upper limit predicted by the Krieger-
the G10 ratio for pitch/coke mixtures are larger than the upper limit predicted by the Krieger-Dougherty
relation. Thus,relation.
Dougherty it can be concluded
Thus, it can bethat these models
concluded maymodels
that these not bemay appropriate to study the
not be appropriate rheological
to study the
parameters of particles mixed with non-Newtonian viscoelastic
rheological parameters of particles mixed with non-Newtonian viscoelastic fluids. fluids.

Figure
Figure 14.14. Effect
Effect ofof cokeparticle
coke particlecontent
contenton
onGG/G ratio (G
1 /G1 0 ratio (G1 and
andG
G100are
arerespectively
respectivelythe storage
the moduli
storage moduli
0
of the mixture and pitch).
of the mixture and pitch).
 Figure
Materials 2016,14. Effect of coke particle content on G/G0 ratio (G and G0 are respectively the storage moduli
9, 334 11 of 13
of the mixture
Figure and
14. Effect pitch).
of coke particle content on G/G0 ratio (G and G0 are respectively the storage moduli
of the mixture and pitch).

Figure 15.
Figure 15. Effect
Effect of
of coke
coke particle
particle content
content on
on G
G/G ratio (G
2 G2 0 ratio (G and G
2 and G are respectively
2 0 are respectively the
the loss
loss moduli
moduli
0 0
of
of the mixture
the mixture
Figure and
and
15. Effect pitch).
ofpitch).
coke particle content on G/G0 ratio (G and G0 are respectively the loss moduli
of the mixture and pitch).

Figure 16. Effect of fine coke particle content on G/G0 ratio. Experimental data are compared with
the Hashin-Shtrikman
Figure 16. Effect
Effectof fineand
offine Krieger-Dougherty
coke particlecontent
content equations.
Figure 16. coke particle onon
G1G/G 0 ratio. Experimental data are compared with
/G1 0 ratio. Experimental data are compared with the
the Hashin-Shtrikman and Krieger-Dougherty
Hashin-Shtrikman and Krieger-Dougherty equations. equations.

5. Conclusions
A cohesive viscoelastic contact model based on Burgers model was implemented in YADE.
The rheological behavior of pitch was measured at 150 C using a dynamic shear rheometer. The obtained
data were then used to calibrate the DEM model parameters for pitch. DSR test of pitch was then
simulated by a three-dimensional DEM model in which pitch is modeled by an assembly of spheres of
radius of 0.08 mm in hcp configuration.
Results confirm that Burgers model is a superior choice to describe the complex rheological
behavior of pitch. The simulation results are in a very good agreement with the experimental data.
However, the storage modulus is over-estimated at high frequencies. It is believed that using a smaller
time-step and not applying the frequency-temperature superposition (which of course results in very
long computation times) can provide a better prediction of phase angle and thus a better prediction of
storage modulus at higher frequencies.
Calibrated DEM model of pitch was then used to predict the rheological properties of coke/pitch
pastes with different content of coke aggregates. Since experimental measurements using a DSR
machine was not possible, the proposed DEM numerical model is of great practical interest.
Results showed that as the content of coke aggregates increases, both storage and loss moduli of
the mixture increase, resulting in an near-constant phase angle. The rise in complex modulus value is
more pronounced at higher frequencies. Comparison of the obtained result with the available literature
shows that the empirical models developed for the case of rigid spheres dispersed in Newtonian fluids
fail to accurately predict the effects of adding coke particles to pitch. More investigations are required
Materials 2016, 9, 334 12 of 13

to elucidate the hydrodynamic interactions between two particles in a viscoelastic suspension and the
rheological behavior of the mixture.

Acknowledgments: The authors gratefully acknowledge the financial support provided by Alcoa Inc., the
Natural Sciences and Engineering Research Council of Canada, and the Centre Qubcois de Recherche et de
Dveloppement de l'Aluminium. A part of the research presented in this article was financed by the Fonds de
recherche du Qubec-Nature et Technologies by the intermediary of the Aluminium Research CentreREGAL.
Author Contributions: Behzad Majidi and Houshang Alamdari conceived and designed the experiments;
Behzad Majidi performed the experiments; Mario Fafard, Donald P. Ziegler and Seyed Mohammad Taghavi
contributed in data analysis; Behzad Majidi wrote the paper and all co-authors commented/corrected it.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Mannwller, U. Anodes for the Aluminium Industry; R & D Carbon Ltd.: Sierre, Switzerland, 1995.
2. Cundall, P.A.; Strack, O.D.L. A discrete numerical model for granular assemblies. Geotechnique 1979, 29,
4765. [CrossRef]
3. Apostolou, K.; Hrymak, A.N. Discrete element simulation of liquid-particle flows. Comput. Chem. Eng. 2008,
32, 841856. [CrossRef]
4. Guo, Y.; Curtis, J.S. Discrete Element Method Simulations for Complex Granular Flows. Annu. Rev.
Fluid Mech. 2015, 47, 2146. [CrossRef]
5. Marketos, G.; Bolton, M.D. Compaction bands simulated in Discrete Element Models. J. Struct. Geol. 2009,
31, 479490. [CrossRef]
6. Harthong, B.; Jrier, J.F.; Dormus, P.; Imbault, D.; Donz, F.V. Modeling of high-density compaction of
granular materials by the Discrete Element Method. Int. J. Solids Struct. 2009, 46, 33573364. [CrossRef]
7. Liu, J.; Sun, Q.; Jina, F.; Liu, Q. Studies on structural and mechanical properties under isostatic compression
with large-scale discrete element simulations. Acta Mech. Solida Sin. 2014, 27, 129136. [CrossRef]
8. Fenga, H.; Pettinari, M.; Hofko, B.; Stang, H. Study of the internal mechanical response of an asphalt mixture
by 3D discrete element modeling. Constr. Build. Mat. 2015, 77, 187196. [CrossRef]
9. Dondi, G.; Simone, A.; Vignali, V.; Manganelli, G. Discrete particle element analysis of aggregate interaction
in granular mixes for asphalt: Combined DEM and experimental study. In 7th RILEM International Conference
on Cracking in Pavements; Springer: Berlin, Germany, 2012.
10. Khattak, M.J.; Khattab, A.; Rizvi, H.R.; Das, S. Imaged-based discrete element modeling of hot mix asphalt
mixtures. Mater. Struct. 2015, 48, 24172430. [CrossRef]
11. Ma, T.; Zhang, D.; Zhang, Y.; Zhao, Y.; Huang, X. Effect of air voids on the high-temperature creep behavior
of asphalt mixture based on three-dimensional discrete element modeling. Mater. Des. 2016, 89, 304313.
[CrossRef]
12. Vignali, V.; Mazzotta, F.; Sangiorgi, C.; Simone, A.; Lantieri, C.; Dondi, G. Rheological and 3D DEM
characterization of potential rutting of cold bituminous mastics. Constr. Build. Mater. 2014, 73, 339349.
[CrossRef]
13. Dondi, G.; Vignali, V.; Pettinari, M.; Mazzotta, F.; Simone, A.; Sangiorgi, C. Modeling the DSR complex shear
modulus of asphalt binder using 3D discrete element approach. Constr. Build. Mater. 2014, 54, 236246.
[CrossRef]
14. Zelelew, H.M. Simulation of the Permanent Deformation of Asphalt Concrete Mixtures Using Discrete
Element Method (DEM). Ph.D. Thesis, Washington State University, Washington, WA, USA, 2008.
15. milauer, V.; Catalano, E.; Chareyre, B.; Dorofeenko, S.; Duriez, J.; Gladky, A.; Kozicki, J.; Modenese, C.;
Scholts, L.; Sibille, L.; et al. Yade Reference Documentation. 2010. Available online: http://yade-dem.org/
doc/ (access on 29 April 2016).
16. Yusoff, N.I.M.; Shaw, M.T.; Airey, G.D. Modelling the linear viscoelastic rheological properties of bituminous
binders. Constr. Build. Mater. 2011, 25, 21712189. [CrossRef]
17. Liu, Y.; You, Z. Accelerated Discrete-Element Modeling of Asphalt-Based Materials with the
Frequency-Temperature Superposition Principle. J. Eng. Mech. 2011, 137, 355365. [CrossRef]
18. Majidi, B.; Azari, K.; Alamdari, H.; Fafard, M.; Ziegler, D. Simulation of vibrated bulk density of anode-grade
coke particles using discrete element method. Powder Technol. 2014, 261, 154160. [CrossRef]
Materials 2016, 9, 334 13 of 13

19. Pasquino, R. Rheology of Viscoelastic Suspensions. Ph.D. Thesis, University of Naples Federico II, Naples,
Italy, 2008.
20. Hashin, Z.; Shtrikman, S. A variational approach to the theory of the elastic behaviour of multiphase
materials. J. Mech. Phys. Solids 1963, 11, 127140. [CrossRef]
21. Krieger, I.M.; Dougherty, T.J. A mechanism for non-Newtonian flow in suspensions of rigid spheres.
Trans. Soc. Rheol. 1959, 3, 137152. [CrossRef]

2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).